Gen AI’s ‘iPhone Moment’: What it Means for Developers
In the year since ChatGPT first launched in November 2022, there’s been speculation about when generative artificial intelligence (gen AI) will have its “iPhone moment” and go truly mainstream. Understanding what this really means and how it will come about is important for developers looking to build successful gen AI applications.
First, it’s helpful to look at what led to the original “iPhone moment.” The device caused a huge splash when it went on sale in the U.S. in mid-2007, but sales in the first year were relatively modest. It was after the App Store debuted in mid-2008 that iPhone sales picked up substantially, and they climbed steadily as more applications were added.
Clearly, the device itself wasn’t the iPhone moment. The breakthrough arrived when people were able to easily discover and download apps that were intuitive and fun to use. This was supported by a marketplace that made those apps easy to discover and enforced the security and governance policies required to protect users and their data.
Gen AI’s ‘iPhone Moment’
The excitement around gen AI is often expressed as the ability to talk to and interact directly with the data, but that’s not entirely true. The people who interact with the data are typically in the back room scripting out Python queries, while everyone else interacts with the applications — like ChatGPT — that are intermediaries between the user and the large language models (LLMs) that power gen AI.
Furthermore, the data doesn’t live on the device you’re using to access the app. It’s stored and processed in the cloud, which means the gen AI era is about much more than just the data and the algorithms. What really brings gen AI to fruition is a combination of the LLMs, plus the cloud data infrastructure, plus the applications.
So, much as the iPhone reached prominence due to the availability of numerous simple applications for discrete tasks, gen AI will become ubiquitous when its capabilities are presented as intuitive applications that are easy to discover and use. In many cases, users won’t even know there’s an AI under the hood; it becomes part of the application, along with natural language processing, AI-assisted search and other embedded functionality.
For developers looking to build successful gen AI applications, that means thinking about more than just what the data can do, but also how users will interact with it. You can build the most sophisticated model in the world, but if it’s surfaced through a complex interface or is difficult to discover and install, it will go to waste.
3 Tips for Building Gen AI Apps
Here are three important elements for developers to consider to build successful gen AI applications.
An Intuitive App Interface
One reason ChatGPT took off so quickly is the bare simplicity of its interface, so think carefully about the user’s level of technical sophistication. Consumers, business executives, data analysts, data engineers and software developers have very different relationships with and understanding of technology, so design an interface that will be perfect for your target users — not necessarily for you.
In many cases, the interface will not be a standalone app. Businesses are already embedding gen AI into existing apps in various ways. Eventually, it’s likely that many apps will feature a gen AI “copilot” that’s ready to answer questions — much like the search bar in apps today.
To build these interfaces, developers need tools that allow them to quickly turn their data, models, and analytic and app functions into interactive apps written in languages such as Python. Streamlit is one option — more than 5,000 LLM apps have already been built on the Streamlit Community Cloud alone — but there are other options, too.
Built-in Data Governance
One reason the App Store became so successful is it provided a secure, tightly controlled environment for building applications. Still today, developers must ask permission to access a user’s contacts, camera or photos, for example. For gen AI applications to really take off, developers need this governance built into their development environment.
That means having an infrastructure layer with the required compliance, security, interoperability and access controls built in, so developers and users see only the data they’re supposed to see. Apps may need to combine data on the provider’s side with data on the consumer side but without revealing the data to either party. This can be achieved with data clean room technology.
This governance should be built into the developer environment from the beginning so that developers aren’t building custom controls for every application. It should be easy for consumers to understand and decide what permissions they want to grant the application. And ideally it will work across different public cloud environments since many organizations are now operating on a multicloud architecture.
A Solid Distribution Mechanism
Target users also need to find and use your applications easily. This is no small matter: Developers need a way to raise awareness of their apps and distribute them through a trusted environment where the intended audience can find and install them easily.
If you’re building consumer apps, that may be through the App Store or the Google Play Store. In a business environment, an app marketplace provides a way to publish apps once and make them available internally to employees or externally to other businesses. Ideally, the marketplace operates across clouds and regions and supports flexible, usage-based business models to monetize the app.
The Gen AI Moment Is Near
Few technologies have captured attention the way generative AI has. IDC predicts that spending on AI products will surpass $500 billion in 2027, driven in part by surging interest in gen AI. If LLMs are to have their iPhone moment, it will happen very soon. To make it happen, focus not just on developing a brilliant model, but on the application that will allow as many people as possible to use it.